Machine Learning Framework for the Sustainable Maintenance of Building Facilities

Abstract

The importance of sustainable building maintenance is growing as part of the Sustainable Building concept. The integration and implementation of new technologies such as the Internet of Things (IoT), smart sensors, and information and communication technology (ICT) into building facilities generate a large amount of data that will be utilized to better manage the sustainable building maintenance and staff. Anomaly prediction models assist facility managers in informing operators to perform scheduled maintenance and visualizing predicted facility anomalies on building information models (BIM). This study proposes a Machine Learning (ML) anomaly prediction model for sustainable building facility maintenance using an IoT sensor network and a BIM model. The suggested framework shows the data management technique of the anomaly prediction model in the 3D building model. The case study demonstrated the framework’s competence to predict anomalies in the heating ventilation air conditioning (HVAC) system. Furthermore, data collected from various simulated conditions of the building facilities was utilized to monitor and forecast anomalies in the 3D model of the fan coil. The faults were then predicted using a classification model, and the results of the models are introduced. Finally, the IoT data from the building facility and the predicted values of the ML models are visualized in the building facility’s BIM model and the real-time monitoring dashboard, respectively.

Understanding the opportunities and challenges of building automation and control systems to support facility management – an extensive literature review

“…Reactive maintenance involves repairing an item to restore its function and is implemented when a defect or malfunction is detected (British Standards Institution, 2018). BACS support facility managers with providing real-time monitoring and notifications of defects, enabling faster response and minimising potential hazards (Villa et al, 2022). However, reactive maintenance can be costly as it requires urgent and unexpected repairs or replacement and it disrupts the optimal operation of the building (Fialho et al, 2022).…”

“…Preventive maintenance, or proactive maintenance, is executed before defects are reported to prevent malfunctions (Villa et al, 2022). Components will be replaced before reaching their projected end of life.…”

“…Reactive maintenance involves repairing an item to restore its function and is implemented when a defect or malfunction is detected (British Standards Institution, 2018). BACS support facility managers with providing real-time monitoring and notifications of defects, enabling faster response and minimising potential hazards (Villa et al. , 2022).…”

A Systematic Review of the Extent to Which BIM Is Integrated into Operation and Maintenance

“…In addition, the framework could benefit from further validations with more case studies so as to enhance the understanding of its proficiency. One unique contribution was by Chew et al [135], who applied the 5G network for smart building and smart facilities management (SFM) in Singapore and proposed a training framework that allowed higher density of internet of things (IoT) device connections, as in [136]. Other integration efforts were considered by Alwan [137], who focused on the refurbishment of housing stock, while Xing et al [138,139] highlighted the interesting integration of Cloud-BIM enabled cyber-physical data and COBie for component reuse.…”

Automated anomaly monitoring and detection system for FCU system

“…Recent studies in the area of AFDD of FCU provided by authors , (Kim et al, 2018), (Schwabacher et al, 2007), (Katipamula et al, 2001) demonstrate that statistical bands, including the control chart approach, pattern recognition techniques, and hypothesis testing on physical models, are typically used to detect faults (Villa et al, 2021). To isolate defects, information flow charts, expert systems, semantic networks, machine learning approaches (Villa et al, 2022), and parameter estimate methods are often applied. To assess faults, heuristic criteria and probabilistic methodologies are applied.…”